A simulation model of the evolution of total sugar content ( C(TS)) in fruit was developed in order to describe the within- and between-genotype variation of C(TS) observed in a peach ( Prunus persica (L.) Batsch) breeding population. The parameter k defines the ratio of carbon used for synthesizing compounds other than sugars for each genotype. Model input variables are dry flesh growth rate and fresh flesh mass of fruit. We estimated k for 137 peach and nectarine genotypes derived from a clone of a wild peach ( Prunus davidiana) by three generations of crosses with commercial nectarine varieties. We tested the predictive quality of the model on independent datasets. Despite an underestimation of the observed C(TS), the correlation between observations and predictions was suitable (0.72). Spearman correlation coefficients between 2001 and 2002 for model input variables and parameter k were higher than for C(TS). None of the three components assimilation supply to the fruit, metabolism, or dilution, seemed to have a greater relative effect on C(TS) variation than the others. Indeed, C(TS) variation seemed to result from the balance between the three components. The interest of this approach, which consists of dissecting traits into components via an ecophysiological model, for breeding strategy and for sugar accumulation studies are discussed.